(click anywhere on page or use keyboard arrow to continute)
Intro to geospatial
- We all work with a variety of data
- Table of trait means for different plants
- Matrix of species presence/absence by site
- Time series of flow volume at a river gauge
- Geospatial data are data too!
- ...but they happen to have:
- info about geographic locations of observations
- info about spatial relationships among them
- ...and therefore we can:
- visualize the data in map form ("layers")
- combine date based on shared location
- Do analyses that exploit spatial relationships
Digital representations of spatial data
- Conceptually, a set of discrete objects
- Simplified representation:
- Points (rain gauges, animal capture locations)
- Lines (continuous transects, streams, roads)
- Polygons (watersheds, patches, species ranges)
- Each 'object' can have multiple attributes
- Topology may be explicit and enforced
- linear networks, polygon adjacency
- Conceptually, a 'field' view of the world
- Simplified representation
- Regular grid, typically one value per cell ("pixel")
- e.g. satellite imagery, digital terrain, interpolated surfaces
- Each pixel usually has a single-valued attribute
- binary, categorical, continues, etc
- Topology is implicitly defined
- Representing the curved Earth in 2D
- Imagine flattening a ping-pong ball
- Printing on paper
- Viewing on screen
- Many calculations are faster/simpler
Projection involves distortion
- Choice of projection involves trade-offs
- Preserve shape? or area?
- Over what map extent?
- Also need to choose coordinate system
- Where is the origin?
- What are the units?
- Data don't need to be projected
- "Geographic projection": latitude-longitude
- Computers aren't constrained to 2D!
- ...but in practice, most software apps are limited
Dealing with projections
- You need to know the projection of your data!
- Changing projections:
- Some software can project "on the fly"
- convenient, but can cause confusion
- Software for reprojection
- Vector: trivial
- Raster: can involve "warping"
Where is lat 34.419, lon -119.699?
- We can't just model earth as a sphere
- to Ellipsoid: squashed, spheroidal shape
- to Geoid: lumpy surface, even at mean sea level
- (for now, ignore real topography)
Datum: Links coordinates on a reference geoid with actual locations on earth
- WGS84, NAD83 both common
- Align to within 1m of each other
- Both are based on GRS80
- Older spatial data may use older datums
- Local datums provide more accuracy in given place
- but probably more than you'll ever need!
- Caution: Interpreting location using the wrong datum can result in positional error of up to 100's of meters!
Data storage formats
- Considerable variation in terms of
- Portability (open standard? proprietary?)
- Compression type and storage size
- Ability to embed metadata
Data storage formats -- vectors
- ESRI Shapefile: multi-file format
- KML: XML-based format popularized by Google
- ESRI Personal GeoDatabase: extends Access databases (ick!)
- CSV??? (at least for point data...)
- many other lesser known
Data storage formats -- rasters
- ASCII grid: text matrix with header rows
- GeoTIFF: binary image format with geo metadata
- Arc/Info binary grid: proprietary binary format, multi-file
- NetCDF: flexible container for array data
- many others: img, bil, grib, lots more